acti-cci draft feedback on the nsf acci task force reports
Post on 24-Feb-2016
40 Views
Preview:
DESCRIPTION
TRANSCRIPT
ACTI-CCI Draft Feedback on the NSF ACCI Task Force Reports
ACTI / Common Solutions Group Joint MeetingStanford University10 January 2012
Emergence of net@edu CCI WG: Campus Cyberinfrastructure
July 2008: Workshop (with CASC) on “… Cyberinfrastructure from Local Campus to National Facilities”
2009: NSF ACCI forms six Task Forcesspring 2011: Tasks Forces Issue ReportsCampus Bridging Task Force very activeACTI CCI WG (re)organized; studies
these reports in context of Campus CIsummer 2011: XSEDE announced as
successor to the TeraGrid
1: Grand Challenge Task ForceFundamental claim: Cyber
Science and Engineering (CS&E) deserves a permanent programmatic place at NSF
Examples:◦ Climate Change and Human Activity◦ Macromolecular structure and complexes◦ Hazard Analysis and Management◦ Virtual product design for Manufacturing
Grand Challenges require …Computational models, methods,
algorithmsHPC (including exascale)SoftwareData and Visualization (critical of
current data infrastructure)Education, Training, and Workforce
DevelopmentGrand Challenge Communities and
Virtual Organizations
Two weaknesses …Fails to emphasize the potential
role for CCI◦The authors are very good leaders in
this area, and thus have excellent access to national centers
◦Pessimistic about impact of large dataFails to emphasize the consequent
need for end-to-end data networking capacity
Impact on CCIThis report is very positive about
CS&E generally and we should absolutely embrace it in its positive statements
The report reminds us that CCI is often off the radar. Better integration of CCI, as with several XSEDE directions and with better bulk data movement, would help.
2: Cyberlearning and Workforce Development Task Force
redesign entire approach to education from scratch exploiting C4: the Continuous Collaborative Computational Cloud that has already formed and continues to develop
What is C4?
Key points of critique …Positive: the emphasis of the
report on impact on broader society (rather than training more MPI coders, say) is very good
Gentle criticisms:◦perhaps too prescriptive, e.g., with
C4 ◦too optimistic about K12 teachers
being prepared to execute the visionBut will stress CCI in some
positive ways!
3: Data and Visualization Task ForcePoints of Agreement to be Stressed:
◦Data Infrastructure / Services: Essential research assets for today¹s science
◦Culture and Sociological Change: Reinforce expectations and encourage data sharing
◦Responsibility for data stewardship is shared among Principal Investigators, research centers, libraries and archives, agencies, and commercial providers
◦Need for proven “best practices”
Gentle CritiqueIn many ways, the Grand Challenge
report makes a more forceful case for Data Infrastructure and for Visualization than does the Data and Visualization Task Force Report
But on those areas that the report touches, it makes solid points from serious data professionals
Note history and makeup of the Task Force
Specific ShortcomingsReport does not identify specific "grand
challenge areas" in either data management or visualization
Need to address the issue of sustaining collections of irreproducible data
Need means to declare / use trust relationships in access control across large time / space dimensions
Specific problems of “big data” (e.g., networking, storage, data analysis) only touched on
Need for better “Big Data Analytics”Visualization explicitly left out
Impact on CCICCI will have a natural role in
◦Data Stewardship◦Big Data Analytics
CCI and the National centers will need to work together to achieve a positive Remote Visualization outcome
We expect XSEDE to contribute here together with CCI
4: High Performance Computing Task ForceVery thin: Three Areas of focus
◦Cyberinfrastructure sustainability◦Exascale computing◦Broader impact
We concur with these positive statements, and note that they are less obvious than they may appear
WeaknessesFocuses solely on the highest end
capability systems◦How might campus layer dovetail?◦Role for high-throughput computing?◦Need for large data flows?
Didn’t address campus level HPC work and bridging◦Will the national centers be
Fortresses or sources of synergy and leadership?
5: Software for Science and Engineering Task ForceSoftware identified as critical success
factorTension between “organically” developed
codes and the need for rigorous software engineering and validation for sustainability
Noted importance of compilers etc. that can leverage modern processors (incl. MICs and GPUs and multi/many-core)
Did not stress issues required for O(10^6) elements
Kinds of Software noted:
◦Operating and grid software◦Compilers and other tools◦Robust applications
Comments:◦History of downplaying importance
and difficulty of software◦Our community is often “in denial”
about how hard this is to do right
6: Campus Bridging Task ForceAt least two major interpretations
of “bridging”:◦bridging “from” campus to “national
CI”◦bridging “among” campus CI and
national CI resources
Recommendations to NSF1. encourage the use of the InCommon
Federation global federated system by using it in services
2. establish a blueprint for a National Cyberinfrastructure
3. create a new program funding high-speed connections from campuses to the nearest landing point for a national network backbone
Recommendations to NSF (cont’d)
4. fund national facilities for data storage and management to support collaboration, scientific workflows, etc.
5. continue research, development, and delivery of new networking technologies
6. fund Cyberinfrastructure activities through careful analyses of needs and outcomes
Recommendations to Higher Ed
1. invest in university, state, and regional cyberinfrastructure –including the needed skilled FTEs
2. have a strategic plan to establish a coherent cyberinfrastructure maximizing value and minimizing global environmental impact
3. adopt criteria for tenure and promotion that reward the range of contributions (data sets, scholarly services, software, etc.)
Recommendations for Commercial Providers
work with the US open research community to reduce barriers to use of their facilities:◦quality of connectivity ◦business model issues such as
transport costs◦policy issues such as the control of
geographic location of data
How Should CCI Respond?Can we (campus CI staff):leverage XSEDE do better alignment with regional and
national CI providers?contribute to controlling cost of
infrastructure and services for research computing?
directly support faculty use of CI at any distance?
promote more interoperability?
Thanks to specific contributors
Grand Challenges (Guy Almes, Jill Gemmill)Cyberlearning (Raj Bose and Sally Jackson)Data and Visualization (Jill Gemmill, Rick
McMullen)HPC (Vijay Agarwala, Curt Hillegas)Software for Science and Engineering (Rick
McMullen, and Jill Gemmill)Campus Bridging (Sally Jackson, Rick
McMullen, Thomas Hauser, Guy Almes)
Questions?
top related